[Feature] SYCL kernel support for Intel GPU

[abhilash1910]

Thanks for creating the project. This is an initial effort to upstream SYCL kernel support for our Intel GPUs. (cxx compiler)

[WoosukKwon]

Hi @abhilash1910, thanks for submitting the great work! Can we chat about the PR? We'd like to know more about the background and your (team's) plan. If you're interested, please reach out to my email: woosuk.kwon[at]berkeley.edu

[abhilash1910]

Thanks @WoosukKwon , I will connect with you .

[jikunshang]

Hi @abhilash1910 @WoosukKwon, Thanks for your great effort to provide this PR. It's really a good start but it seems not able to build or run. Instead of migrating CUDA kernels to SYCL directly, we redesigned Intel XPU support for vllm via dispatching operators to customized SYCL kernels/functions.
Currently, our approach could run an easy e2e example(offline_inference.py) using opt-125m model. The development is in progress, still have a lot to be done. Details can be found in PR #2378. Please take a review, thanks!
cc @bigPYJ1151 @maktukmak

[abhilash1910]

Diverted to #2378 for consolidated porting.

[WoosukKwon]

Closing this PR as we merged #3814